resurrect an old patch from Pablo Neira to remove the untracked objects.
Currently, there are four possible states of an skb wrt. conntrack.
1. No conntrack attached, ct is NULL.
2. Normal (kmem cache allocated) ct attached.
3. a template (kmalloc'd), not in any hash tables at any point in time
4. the 'untracked' conntrack, a percpu nf_conn object, tagged via
IPS_UNTRACKED_BIT in ct->status.
Untracked is supposed to be identical to case 1. It exists only
so users can check
-m conntrack --ctstate UNTRACKED vs.
-m conntrack --ctstate INVALID
e.g. attempts to set connmark on INVALID or UNTRACKED conntracks is
supposed to be a no-op.
Thus currently we need to check
ct == NULL || nf_ct_is_untracked(ct)
in a lot of places in order to avoid altering untracked objects.
The other consequence of the percpu untracked object is that all
-j NOTRACK (and, later, kfree_skb of such skbs) result in an atomic op
(inc/dec the untracked conntracks refcount).
This adds a new kernel-private ctinfo state, IP_CT_UNTRACKED, to
make the distinction instead.
The (few) places that care about packet invalid (ct is NULL) vs.
packet untracked now need to test ct == NULL vs. ctinfo == IP_CT_UNTRACKED,
but all other places can omit the nf_ct_is_untracked() check.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
refcount_t type and corresponding API (see include/linux/refcount.h)
should be used instead of atomic_t when the variable is used as
a reference counter. This allows to avoid accidental
refcounter overflows that might lead to use-after-free
situations.
Signed-off-by: Elena Reshetova <elena.reshetova@intel.com>
Signed-off-by: Hans Liljestrand <ishkamiel@gmail.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David Windsor <dwindsor@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Currently, the ip_vs_dest cache frees ip_vs_dest objects when their
reference count becomes < 0. Aside from not being semantically sound,
this is problematic for the new type refcount_t, which will be introduced
shortly in a separate patch. refcount_t is the new kernel type for
holding reference counts, and provides overflow protection and a
constrained interface relative to atomic_t (the type currently being
used for kernel reference counts).
Per Julian Anastasov: "The problem is that dest_trash currently holds
deleted dests (unlinked from RCU lists) with refcnt=0." Changing
dest_trash to hold dest with refcnt=1 will allow us to free ip_vs_dest
structs when their refcnt=0, in ip_vs_dest_put_and_free().
Signed-off-by: David Windsor <dwindsor@gmail.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Add a helper to assign a nf_conn entry and the ctinfo bits to an sk_buff.
This avoids changing code in followup patch that merges skb->nfct and
skb->nfctinfo into skb->_nfct.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Next patch makes direct skb->nfct access illegal, reduce noise
in next patch by using accessors we already have.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Previous patch that introduced handling of outgoing packets in SIP
persistent-engine did not call ip_vs_check_template() in case packet was
matching a connection template. Assumption was that real-server was
healthy, since it was sending a packet just in that moment.
There are however real-server fault conditions requiring that association
between call-id and real-server (represented by connection template)
gets updated. Here is an example of the sequence of events:
1) RS1 is a back2back user agent that handled call-id1 and call-id2
2) RS1 is down and was marked as unavailable
3) new message from outside comes to IPVS with call-id1
4) IPVS reschedules the message to RS2, which becomes new call handler
5) RS2 forwards the message outside, translating call-id1 to call-id2
6) inside pe->conn_out() IPVS matches call-id2 with existing template
7) IPVS does not change association call-id2 <-> RS1
8) new message comes from client with call-id2
9) IPVS reschedules the message to a real-server potentially different
from RS2, which is now the correct destination
This patch introduces ip_vs_check_template() call in the handling of
outgoing packets for SIP-pe. And also introduces a second optional
argument for ip_vs_check_template() that allows to check if dest
associated to a connection template is the same dest that was identified
as the source of the packet. This is to change the real-server bound to a
particular call-id independently from its availability status: the idea
is that it's more reliable, for in->out direction (where internal
network can be considered trusted), to always associate a call-id with
the last real-server that used it in one of its messages. Think about
above sequence of events where, just after step 5, RS1 returns instead
to be available.
Comparison of dests is done by simply comparing pointers to struct
ip_vs_dest; there should be no cases where struct ip_vs_dest keeps its
memory address, but represent a different real-server in terms of
ip-address / port.
Fixes: 39b9722315 ("ipvs: handle connections started by real-servers")
Signed-off-by: Marco Angaroni <marcoangaroni@gmail.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
When using LVS-NAT and SIP persistence-egine over UDP, the following
limitations are present with current implementation:
1) To actually have load-balancing based on Call-ID header, you need to
use one-packet-scheduling mode. But with one-packet-scheduling the
connection is deleted just after packet is forwarded, so SIP responses
coming from real-servers do not match any connection and SNAT is
not applied.
2) If you do not use "-o" option, IPVS behaves as normal UDP load
balancer, so different SIP calls (each one identified by a different
Call-ID) coming from the same ip-address/port go to the same
real-server. So basically you don’t have load-balancing based on
Call-ID as intended.
3) Call-ID is not learned when a new SIP call is started by a real-server
(inside-to-outside direction), but only in the outside-to-inside
direction. This would be a general problem for all SIP servers acting
as Back2BackUserAgent.
This patch aims to solve problems 1) and 3) while keeping OPS mode
mandatory for SIP-UDP, so that 2) is not a problem anymore.
The basic mechanism implemented is to make packets, that do not match any
existent connection but come from real-servers, create new connections
instead of let them pass without any effect.
When such packets pass through ip_vs_out(), if their source ip address and
source port match a configured real-server, a new connection is
automatically created in the same way as it would have happened if the
packet had come from outside-to-inside direction. A new connection template
is created too if the virtual-service is persistent and there is no
matching connection template found. The new connection automatically
created, if the service had "-o" option, is an OPS connection that lasts
only the time to forward the packet, just like it happens on the
ingress side.
The main part of this mechanism is implemented inside a persistent-engine
specific callback (at the moment only SIP persistent engine exists) and
is triggered only for UDP packets, since connection oriented protocols, by
using different set of ports (typically ephemeral ports) to open new
outgoing connections, should not need this feature.
The following requisites are needed for automatic connection creation; if
any is missing the packet simply goes the same way as before.
a) virtual-service is not fwmark based (this is because fwmark services
do not store address and port of the virtual-service, required to
build the connection data).
b) virtual-service and real-servers must not have been configured with
omitted port (this is again to have all data to create the connection).
Signed-off-by: Marco Angaroni <marcoangaroni@gmail.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Jiri Bohac is reporting for a problem where the attempt
to reschedule existing connection to another real server
needs proper redirect for the conntrack used by the IPVS
connection. For example, when IPVS connection is created
to NAT-ed real server we alter the reply direction of
conntrack. If we later decide to select different real
server we can not alter again the conntrack. And if we
expire the old connection, the new connection is left
without conntrack.
So, the only way to redirect both the IPVS connection and
the Netfilter's conntrack is to drop the SYN packet that
hits existing connection, to wait for the next jiffie
to expire the old connection and its conntrack and to rely
on client's retransmission to create new connection as
usually.
Jiri Bohac provided a fix that drops all SYNs on rescheduling,
I extended his patch to do such drops only for connections
that use conntrack. Here is the original report from Jiri Bohac:
Since commit dc7b3eb900 ("ipvs: Fix reuse connection if real server
is dead"), new connections to dead servers are redistributed
immediately to new servers. The old connection is expired using
ip_vs_conn_expire_now() which sets the connection timer to expire
immediately.
However, before the timer callback, ip_vs_conn_expire(), is run
to clean the connection's conntrack entry, the new redistributed
connection may already be established and its conntrack removed
instead.
Fix this by dropping the first packet of the new connection
instead, like we do when the destination server is not available.
The timer will have deleted the old conntrack entry long before
the first packet of the new connection is retransmitted.
Fixes: dc7b3eb900 ("ipvs: Fix reuse connection if real server is dead")
Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
This function adds no real value and it obscures what the code is doing.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
This hack has no more users so remove it.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
The argument is unnecessary and in practice confusing,
and has caused the callers to do all manner of silly things.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
With sysctl_cache_bypass now a compile time constant the compiler can
figue out that it can elimiate all of the code that depends on
sysctl_cache_bypass being true.
Also remove the duplicate computation of net previously necessitated
by #ifdef CONFIG_SYSCTL
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
This moves the hack "net_ipvs(skb_net(skb))" up one level where it
will be easier to remove.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Move the hack of relying on "net_ipvs(skb_net(skb))" to derive the
ipvs up a layer.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Stop relying on "net_ipvs(skb_net(skb))" to derive the ipvs as
skb_net is a hack.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Also move the tests for net_ipvs being NULL into __ip_vs_ftp_init
and __ip_vs_ftp_exit. The only places where they possibly make
sense.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
In practice struct netns_ipvs is as meaningful as struct net and more
useful as it holds the ipvs specific data. So store a pointer to
struct netns_ipvs.
Update the accesses of param->net to access param->ipvs->net instead.
In functions where we are searching for an svc and filtering by net
filter by ipvs instead.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
ipvs is what is actually desired so change the parameter and the modify
the callers to pass struct netns_ipvs.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
In practice struct netns_ipvs is as meaningful as struct net and more
useful as it holds the ipvs specific data. So store a pointer to
struct netns_ipvs.
Update the accesses of param->net to access param->ipvs->net instead.
When lookup up struct ip_vs_conn in a hash table replace comparisons
of cp->net with comparisons of cp->ipvs which is possible
now that ipvs is present in ip_vs_conn_param.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
In practice struct netns_ipvs is as meaningful as struct net and more
useful as it holds the ipvs specific data. So store a pointer to
struct netns_ipvs.
Update the accesses of conn->net to access conn->ipvs->net instead.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
This is a way to avoid nasty routing loops when multiple ipvs instances can
forward to eachother.
Signed-off-by: Alex Gartrell <agartrell@fb.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
This sysctl will be used to enable the scheduling of icmp packets.
Signed-off-by: Alex Gartrell <agartrell@fb.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
No longer necessary since the information is included in the ip_vs_iphdr
itself.
Signed-off-by: Alex Gartrell <agartrell@fb.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
These flags contain information like whether or not the addresses are
inverted or from icmp. The first will allow us to drop an inverse param
all over the place, and the second will later be useful in scheduling icmp.
Signed-off-by: Alex Gartrell <agartrell@fb.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
This removes some duplicated code and makes the ICMPv6 path look more like
the ICMP path.
Signed-off-by: Alex Gartrell <agartrell@fb.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
- mcast_group: configure the multicast address, now IPv6
is supported too
- mcast_port: configure the multicast port
- mcast_ttl: configure the multicast TTL/HOP_LIMIT
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Allow setups with large MTU to send large sync packets by
adding sync_maxlen parameter. The default value is now based
on MTU but no more than 1500 for compatibility reasons.
To avoid problems if MTU changes allow fragmentation by
sending packets with DF=0. Problem reported by Dan Carpenter.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Having to say
> #ifdef CONFIG_NET_NS
> struct net *net;
> #endif
in structures is a little bit wordy and a little bit error prone.
Instead it is possible to say:
> typedef struct {
> #ifdef CONFIG_NET_NS
> struct net *net;
> #endif
> } possible_net_t;
And then in a header say:
> possible_net_t net;
Which is cleaner and easier to use and easier to test, as the
possible_net_t is always there no matter what the compile options.
Further this allows read_pnet and write_pnet to be functions in all
cases which is better at catching typos.
This change adds possible_net_t, updates the definitions of read_pnet
and write_pnet, updates optional struct net * variables that
write_pnet uses on to have the type possible_net_t, and finally fixes
up the b0rked users of read_pnet and write_pnet.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when TCP/SCTP port reusing happens, IPVS will find the old
entry and use it for the new one, behaving like a forced persistence.
But if you consider a cluster with a heavy load of small connections,
such reuse will happen often and may lead to a not optimal load
balancing and might prevent a new node from getting a fair load.
This patch introduces a new sysctl, conn_reuse_mode, that allows
controlling how to proceed when port reuse is detected. The default
value will allow rescheduling of new connections only if the old entry
was in TIME_WAIT state for TCP or CLOSED for SCTP.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
IPVS stats are limited to 2^(32-10) conns/s and packets/s,
2^(32-5) bytes/s. It is time to use 64 bits:
* Change all conn/packet kernel counters to 64-bit and update
them in u64_stats_update_{begin,end} section
* In kernel use struct ip_vs_kstats instead of the user-space
struct ip_vs_stats_user and use new func ip_vs_export_stats_user
to export it to sockopt users to preserve compatibility with
32-bit values
* Rename cpu counters "ustats" to "cnt"
* To netlink users provide additionally 64-bit stats:
IPVS_SVC_ATTR_STATS64 and IPVS_DEST_ATTR_STATS64. Old stats
remain for old binaries.
* We can use ip_vs_copy_stats in ip_vs_stats_percpu_show
Thanks to Chris Caputo for providing initial patch for ip_vs_est.c
Signed-off-by: Chris Caputo <ccaputo@alt.net>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>